ByteDance-Seed

/

AHN-Mamba2-for-Qwen-2.5-Instruct-3B

like 5

Follow

ByteDance Seed 2.66k

This PR significantly enhances the model card by:

• Adding an explicit link to the Hugging Face paper page (Artificial Hippocampus Networks for Efficient Long-Context Modeling) in the introduction for better discoverability, while retaining the existing arXiv badge.
• Including a Hugging Face Models collection badge from the GitHub README, providing an easy way to explore related models on the Hub.
• Integrating the comprehensive Usage section from the GitHub repository, which includes:
  • Detailed installation instructions.
  • Code snippets for inference.
  • Guidance and key arguments for training.
  • Debugging information.
    This provides users with clear, actionable steps to work with the model.
• Adding the News section from the GitHub repository to keep users informed of updates.
• Incorporating the Acknowledgments section from the GitHub repository, giving credit to contributors and dependencies.

The existing metadata, including pipeline_tag: text-generation, library_name: transformers, and the license details (license: other, license_name: qwen-research), remains unchanged as it is accurate and consistent with the provided evidence and the majority decision from colleagues. The library_name: transformers is supported by the config.json and the model's architecture, and while specific transformers usage snippets are not added (to avoid making up code), the inference script provided in the GitHub README likely leverages it.

These improvements aim to make the model card a more complete and user-friendly resource on the Hugging Face Hub.